Transformer Purchase Price Prediction Method Based on Long Short Term Memory Model with Complex Factor Input

Abstract

Traditional Long Short Term Memory (LSTM) model can effectively improve the price prediction problem, but the LSTM model only accepts a single element input, and the power equipment price prediction involves multiple factors. Based on this, this paper proposes a multi-factor input-based LSTM model for transformer purchase price prediction, firstly, by changing the mapping structure of input layer, loop layer and output layer in the LSTM model, and then reasonably introducing the loop dropout restriction mechanism to ensure that the model structure is efficiently adapted to multi-factor input and low data training time. Finally, experimental simulations are conducted on a real price data set, and the validation results show that this algorithm can efficiently improve the price prediction accuracy and model robustness compared with traditional models.